JP7157245B2 - File management device, file management method, and program - Google Patents

Description

The present invention relates to a file management device, file management method, and program.

For example, Patent Document 1 discloses classification rule registration means for registering a combination of a first rule and a second rule applied to the first rule as a classification rule for classifying documents, and the classification rule registration means A document processing apparatus having a classification rule integration means for integrating a plurality of different classification rules registered by the different classification rules by eliminating duplicate contents among the plurality of different classification rules.

Further, Patent Document 2 describes a step of storing a new data item and a category of the new data item in a storage device, and data stored in a correct data storage unit storing the data item and the category of the data item. a step of extracting a feature pattern including a condition including a feature element of the new data item stored in the storage device and a corresponding category from the storage device and storing the extracted feature pattern in a feature pattern storage unit; The feature patterns obtained are grouped into a first set that matches the category of the new data item stored in the storage device and a second set that does not match the category of the new data item, and the result of grouping is stored in a group data storage unit. A computer-implemented classification rule creation support method is disclosed, comprising:

Further, in Patent Document 3, a storage means, an input means for inputting document image data representing a document, and a layout analysis of the document image data input to the input means to perform a layout analysis of the document represented by the document image data. determining means for performing character analysis on the document image data input to the input means and determining the attribute of each description item of the document represented by the document image data; and the layout specified by the specifying means. and generation means for specifying a hierarchical structure between entry items based on the attribute of each entry item determined by the determination means and generating rule data representing the hierarchical structure; A document processing device is disclosed comprising writing means for writing to a storage means.

JP 2013-251610 JP 2007-052744 JP 2007-052615

An object of the present invention is to provide a file management device capable of appropriately classifying data files.

A file management apparatus according to the present invention includes a common feature extraction unit for extracting features common to a plurality of data files to which the same tag is attached, and a feature extracted by the common feature extraction unit. , a rule storage unit that associates tags assigned to these data files with each other and stores them as assignment rules; and a tag assigning unit that assigns

Preferably, the tagging unit searches the newly input data file for features registered as the tagging rules in the rule storage unit, and if any of the features is found, Applies associated tags to newly input data files.

Preferably, the tagging unit, when part of the feature registered as the tagging rule is found from a newly input data file, proposes a tag associated with the feature to the user, A tag is added according to the user's operation.

Preferably, when the proposed tag is adopted by the user, it further has a rule updating unit that updates the grant rule so that the characteristics of the newly input data file match the characteristics of the grant rule.

Preferably, it further comprises a rule updating unit that updates the granting rule so that the features of the newly input data file do not match the features of the granting rule when the proposed tag is not adopted by the user.

Preferably, the common feature extraction unit extracts at least one of a character string, date, image size, and number of colors used in the image as the feature.

Preferably, the given rule stored in the rule storage section includes a plurality of determination elements, and the rule update section selects appearance frequency, recency, and appearance from features common to a plurality of data files. Based on at least one of the locations and the peculiarity, a feature is selected to be registered as a determining factor of the application rule.

A file management method according to the present invention comprises a common feature extraction step of extracting features common to a plurality of data files to which the same tag is assigned, and features extracted by the common feature extraction step. , a rule storage step of correlating the tags assigned to these data files with each other and storing the tags as assignment rules; and a tagging step of giving

A program according to the present invention comprises a common feature extraction step of extracting features common to a plurality of data files to which the same tag is assigned, features extracted by the common feature extraction step, and A rule storing step for storing the tag assigned to the data file of the above as an assigning rule in association with each other, and assigning a tag to the newly input data file based on the assigning rule stored in the rule storing step and a tagging step.

Data files can be properly categorized.

1 is a diagram illustrating the overall configuration of a file management system 1; FIG. 3 is a diagram illustrating a hardware configuration of a file management device 5; FIG. 3 is a diagram illustrating a functional configuration of a file management device 5; FIG. (a) is a table for explaining the tagging rules, (b) is a table for explaining the tagging rules for the tag "estimate", and (c) is a table for explaining the tagging rules extracted by the common feature extraction unit 502. FIG. 4 is a diagram illustrating features; (a) is a table showing examples of keywords and position information of the keywords, (b) is a diagram showing examples of dates and position information of the dates, and (c) is an example of image sizes and the number of colors used. It is a table. 4 is a flowchart for explaining tagging rule registration and update processing (S10) by the file management device 5. FIG. It is an example of a screen of a tagging operation by a user. (a) is an invocation example 1 of a tagging rule customization screen, and (b) is an example of a tagging rule customization screen by a user. (a) is a calling example 2 of the tagging rule customization screen, and (b) is a calling example 3 of the tagging rule customization screen. 4 is a flowchart for explaining tagging and tag proposal processing (S20) by the file management device 5; FIG. 10 is a flow chart illustrating a tagging rule updating process (S30) according to a user's response to a tagging proposal; FIG. FIG. 11 is a diagram illustrating a user response operation screen for tagging alternatives; (a) is a diagram exemplifying the tagging rule of "invoice (2018)" and condition candidates for document C, (b) is a table showing keyword scores of the condition candidates, and (C) is , which is an example of an updated tagging rule. (a) is a diagram showing the features of the tagging rule for "bill" and the document D, (b) is a table showing the scores of keywords for condition candidates, and (c) is a tagging rule is an example of updating. (a) is a diagram showing the tagging rules of "Company AA_bill" and the characteristics of the newly tagged document G, which is the invoice of Company BB; FIG. 10 is a diagram illustrating tagging rules for "Kaki", (c) is a table showing scores of keywords of condition candidates, and (d) is an example of updating tagging rule data.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating the overall configuration of a file management system 1. As shown in FIG.
As illustrated in FIG. 1, the file management system 1 includes a plurality of scanners 3a, 3b, 3c, a file management device 5, and a user terminal 7, which are interconnected via a network 9. FIG. The scanner 3a, the scanner 3b, and the scanner 3c are collectively called the scanner 3, and the user terminal 7a and the user terminal 7b are collectively called the user terminal 7. FIG.
The scanner 3 is an optical reading device, and transmits acquired image data to the file management device 5 .
The file management device 5 is a computer terminal, and attaches tags for classifying the image data received from the scanner 3 to the image data. Specifically, the file management device 5 holds tagging rules, which are rules for tagging, and assigns suitable tags to data files based on the tagging rules and the characteristics of data files obtained by OCR processing image data. Grant to. Further, the file management device 5 generates and updates tagging rules according to user's operations. Note that the tagging rule is an example of an assignment rule according to the present invention.
A user terminal 7 is a computer terminal operated by a user, and displays a user interface provided by the file management device 5 .

FIG. 2 is a diagram illustrating the hardware configuration of the file management device 5. As shown in FIG.
As illustrated in FIG. 2, the file management device 5 has a CPU 200, a memory 202, an HDD 204, a network interface 206 (network IF 206), a display device 208, and an input device 210. These components are connected via a bus 212. connected to each other.
CPU 200 is, for example, a central processing unit.
The memory 202 is, for example, a volatile memory and functions as a main memory.
The HDD 204 is, for example, a hard disk drive device, and stores computer programs (eg, the file management program 50 in FIG. 3) and other data files (eg, the tagging rule database 600 in FIG. 3) as a non-volatile recording device. .
The network IF 206 is an interface for wired or wireless communication, and implements communication in the internal network 9, for example.
The display device 208 is, for example, a liquid crystal display.
Input device 210 is, for example, a keyboard and mouse.

FIG. 3 is a diagram illustrating the functional configuration of the file management device 5. As shown in FIG.
As illustrated in FIG. 3, the file management device 5 of this example is installed with a file management program 50 and configured with a tagging rule database 600 (tagging rule DB 600).
The file management program 50 has an acquisition unit 500 , a common feature extraction unit 502 , a matching unit 504 , a score calculation unit 506 , a tagging unit 508 and a rule updating unit 510 .
Part or all of the file management program 50 may be realized by hardware such as ASIC, or may be realized by partially borrowing functions of an OS (Operating System).
In the file management program 50 , an acquisition unit 500 acquires image data read by the scanner 3 .
A common feature extraction unit 502 extracts features common to a plurality of data files to which the same tag is assigned. For example, a data file refers to image data subjected to OCR (Optical Character Recognition) processing. Specifically, the common feature extraction unit 502 performs OCR processing on the image data acquired from the acquisition unit 500, and extracts features of the data file based on the OCR processing result. More specifically, the common feature extraction unit 502 extracts at least one of a character string described in the data file, the date, the image size of the data file, and the number of colors used in the image of the data file as a feature. do.

Here, the tagging rule DB 600 will be explained.
The tagging rule DB 600 associates the features extracted by the common feature extraction unit 502 with the tags given to these data files and stores them as tagging rules. Specifically, the tagging rule DB 600 stores tag names, and “condition candidates” and “conditions” associated with the tag names. A "condition" is a constraint for assigning a "tag name" associated with a data file, and is a common feature extracted from multiple data files assigned the same tag. Similarly, "candidate conditions" are common features extracted from a plurality of data files, and "conditions" are selected from among the "candidate conditions." The tagging rule DB 600 is an example of a rule storage section according to the present invention.

The collation unit 504 collates the newly input data file with features registered as tagging rules. A newly input data file is a data file obtained by OCR processing the image data acquired by the acquisition unit 500 . Specifically, the matching unit 504 determines the degree of matching between the data file OCR-processed by the common feature extraction unit 502 and the features registered in the tagging rules stored in the tagging rule DB 600 .
The score calculation unit 506 calculates a score that is a determination factor for each “condition candidate”, and selects a “condition” from among the “condition candidates” having a score equal to or greater than a threshold. Specifically, the score calculation unit 506 calculates the scores of appearance frequency, recency, appearance position, and uniqueness of each “candidate condition”, weights the scores, and sums them up. Calculate superiority.

The tagging unit 508 tags the newly input data file based on the tagging rules stored in the tagging rule DB 600 .
Specifically, the tagging unit 508 searches the newly input data file for features registered as tagging rules, and if any feature is found, the tagging unit 508 associates with this feature. Applies tags to newly entered data files.
More specifically, when some of the features registered as tagging rules are found in a newly input data file, the tagging unit 508 provides the user with tags associated with the features. Suggestions are made and tags are given according to the user's operation. A case where a part of the feature is found from a newly input data file is a case where the matching rate between the feature extracted by the common feature extraction unit 502 and the feature of the tagging rule is 50% to 99%. Say.

The rule updating unit 510 generates and updates tagging rules. Specifically, the rule updating unit 510 uses at least one of appearance frequency, recency, and appearance position among the features common to a plurality of data files, and uniqueness as a judgment element of the tagging rule. Select the features to be registered. More specifically, the rule updating unit 510 selects “conditions” from “candidate conditions” for which the total score calculated by the score calculating unit 506 is higher than the threshold, and updates the tagging rule.
More specifically, the rule updating unit 510 updates the rule when the matching unit 504 determines that the newly input data file matches part of the characteristics registered as the tagging rule. , if the user adopts the suggested tag, update the tagging rule so that the newly input data file matches the features registered as the tagging rule.
Further, the rule updating unit 510 updates the proposed user when the collating unit 504 determines that the newly input data file matches part of the characteristics registered as the tagging rule. update the tagging rules so that the newly input data file does not match the features registered as the tagging rules when rejecting new tags.

Next, tagging rules stored in the tagging rule DB 600 will be described.
FIG. 4(a) is a table for explaining the tagging rules, (b) is a table for explaining the tagging rules for the tag "estimate", and (c) is a table for explaining the tagging rules extracted by the common feature extraction unit 502. FIG. 10 illustrates the features that are used;
As illustrated in FIG. 4(a), a tagging rule has a "tag name" and a "condition" associated with the "tag name". "Tag name" is a name for classifying data files. A "condition" is a constraint for assigning a "tag name" and is a common feature extracted from data files assigned the same tag. If the data file fetched next time satisfies the 'condition', the file management device 5 automatically assigns the 'tag name' associated with the 'condition' to the fetched data file.

Further, the "condition" is selected from among the "candidate conditions". A “condition candidate” is a feature extracted from a data file by the common feature extraction unit 502 . Specifically, as exemplified in FIG. 4B, the tag "estimate" is associated with "candidate conditions" and "conditions", and the "candidate conditions" is associated with the "conditions" of tagging. are features extracted from the data file that were not included in the , but are saved as candidate "conditions" for when updating the tagging rules.
More specifically, the common feature extraction unit 502 extracts "condition candidates" for each item from the data file illustrated in FIG. 4(c), as illustrated in FIG. 4(b). The score calculation unit 506 calculates a score representing the superiority of each condition candidate, and the rule update unit 510 selects “condition candidates” having a score equal to or higher than the threshold based on the total score of each “condition candidate”. Select "Conditions".

FIG. 5 is a diagram illustrating condition candidates extracted from a data file.
As exemplified in FIG. 5, “candidate conditions” are features extracted from the data file by the common feature extraction unit 502, and are the vertical and horizontal lengths of character strings and images. Specifically, the “candidate conditions” are words (keywords) in the text of the data file, document date values, image length and width values, and the like.
As for the keyword, as illustrated in FIG. 5A, the common feature extraction unit 502 records the character string of the keyword in the data file and the position information describing the character string as the feature. Specifically, the common feature extraction unit 502 performs morphological analysis of the data file based on the result of the OCR processing, and uses the divided words as keyword condition candidates. Among the keyword condition candidates, the common feature extraction unit 502 sets up to five cases whose scores satisfy a specific condition as keyword conditions.

As for the document date, as illustrated in FIG. 5B, the common feature extraction unit 502 decomposes the date in the data file into elements of year/month/day, and records the position information in which the date is described as a feature. do. When the date is written in the document, the common feature extraction unit 502 selects the elements of year, month, day, and day of the week as document date condition candidates. Among the document date condition candidates, the common feature extraction unit 502 selects, as the document date condition, at most one item from each element whose score satisfies a specific condition.
Regarding the image size, as illustrated in FIG. 5C, the common feature extraction unit 502 sets the length and width of the data file as image size condition candidates, and among the image size condition candidates, the score is specified as a condition A maximum of one image size condition is set for each element that satisfies
In addition to the keywords, document date values, and image lengths and widths, the rule updating unit 510 also supports “format,” “business card or receipt attribute values (company name or address),” and “image A tagging rule may be created using "color" as a candidate condition and matching or similar to these as a condition.

The tagging unit 508 tags the data files that satisfy the "conditions". Specifically, the tagging unit 508 adds tags when the characteristics of the data file match the keywords of the tagging rule, when the date of the document is similar, and when the vertical and horizontal lengths of the image are similar.
For example, the keyword condition is satisfied by having a specific character string written in the data file. A similar condition for document dates is met by having certain features of the date listed in the data file. The similarity condition of image length and width is satisfied by having a certain feature of image length and width.

Next, a method for calculating the score of the condition candidate will be described.
Each candidate condition has a score for frequency of occurrence, recency, location of occurrence, and uniqueness. Each score increases or decreases between 0-10.
A frequency of occurrence score is calculated based on how many data files a feature appears in all tagged data files. The score of appearance frequency becomes higher as the number of appearances increases. Also, the score is 10 if the feature is common to all tagged data files.
A recency score is calculated based on whether a feature applies to a recently entered data file. The initial value for recency score is the maximum value (10). Also, the recency score is reduced when the feature does not fit the added data file.
The score of the appearance position is calculated based on whether or not it is described in a close position on the data file. The score of the appearance position is the maximum value (10) at the same location, and the score of the appearance position decreases as the position is farther away.

A uniqueness score is calculated based on whether or not the feature is unique to the tagging rule. When a tagging proposal is made to a data file corresponding to the tagging rule and the user rejects the proposal, the score calculation unit 506 determines that the tagging rule exists and the corresponding data file has adds the uniqueness scores of non-existing "conditions" and "candidate conditions".
The score calculation unit 506 calculates the scores of appearance frequency, recency, appearance position, and peculiarity, weights and adds up the scores, and calculates the superiority of the “condition candidate”. The score calculation unit 506 calculates the total score using the formula “total score=α×appearance frequency score+β×recency score+γ×appearance position score+δ×uniqueness score”. The rule update unit 510 selects “conditions” from condition candidates whose total score calculated by the score calculation unit 506 is higher than the threshold.

FIG. 6 is a flowchart for explaining tagging rule registration and update processing (S10) by the file management device 5. As shown in FIG. FIG. 6 describes the registration and update of the tagging rule when the user attaches the tag "A" to the data file.
As illustrated in FIG. 6, at step 100 (S100), the rule updating unit 510 detects that a tag has been added to the data file by the user's tagging operation. Specifically, as exemplified in FIG. 7, the user selects a tag or inputs a new tag name on the tagging operation screen, and performs a tagging operation (adding tag "A") to the data file. , and the rule updating unit 510 detects the tagging operation by the user.
At step 105 (S105), the matching unit 504 searches the tagging rule DB 600 to see if there is a tagging rule for the tag "A". If the tagging rule exists, the matching unit 504 proceeds to S135, and if the tagging rule does not exist, the matching unit 504 proceeds to S110.

At step 110 (S110), if there are two or more data files to which the tag "A" has been assigned by the user, which are retrieved by the collation unit 504, the collation unit 504 proceeds to S115. If there is only one data file to which the tag "A" is assigned by the user, the rule updating unit 510 terminates the tagging rule registration and update process (S10). If there is only one data file with the tag "A", no tagging rule is generated because common features in data files with the same tag "A" cannot be extracted.
At step 115 (S115), the common feature extraction unit 502 extracts the features of the data files to which the tag "A" is attached. Specifically, the common feature extraction unit 502 extracts at least one of the character string, date, image size, and number of colors used in the image from the data file.

In step 120 (S120), when the common feature extraction unit 502 has extracted the features of all the data files to which the tag "A" is attached, the tagging rule registration and update process (S10) proceeds to S125. However, if the features of all data files have not been extracted, the tagging rule registration and update process (S10) proceeds to S115.
At step 125 (S125), the common feature extraction unit 502 extracts features common to all data files to which the tag "A" is assigned as "candidate conditions".
At step 130 (S130), the score calculation unit 506 calculates the score of appearance frequency, recency, appearance position, and specificity of each “condition candidate”, and a total score. The rule updating unit 510 selects “condition candidates” whose total score is high and whose score is equal to or greater than the threshold as “conditions”, generates a tagging rule for the tag “A”, and registers it in the tagging rule DB 600. .

In step 135 (S135), if there is a tagging rule for tag "A", the matching unit 504 acquires the tagging rule for tag "A".
At step 140 (S140), the common feature extraction unit 502 extracts the features of the data files to which the tag "A" has been assigned by the user. Specifically, the common feature extraction unit 502 acquires at least one of the character string, date, image size, and number of colors used in the image of the data file.
At step 145 (S145), the rule updating unit 510 deletes, from among the "conditions" of the tagging rules acquired by the matching unit 504 at S135, conditions that do not correspond to the features extracted by the common feature extraction unit 502 at S140. do. Furthermore, in S135, the score calculation unit 506 adds the “candidate conditions” of the tagging rule acquired by the matching unit 504, and recalculates the score of each “candidate condition” acquired in S140. When a fixed condition is set by customization by the user, the condition candidate for which the fixed condition is set is selected as the "condition" regardless of the score value. Furthermore, the rule update unit 510 additionally selects, as a “condition”, a condition candidate whose total score is high and whose score is equal to or greater than the threshold value from among the other condition candidates.
At step 150 (S150), the rule updating section 510 updates the tagging rule by replacing the "condition" of the tagging rule for the tag "A" with the selected new "condition". By replacing it with a new "condition", a tagging rule with a higher matching rate can be generated.

Next, customization of tagging rules by the user will be described.
The user can call the tagging rule customization screen at any time to check and customize the tagging rule. Specifically, as exemplified in FIG. 8A, a menu of "confirmation of tagging rules" is displayed from the tag "statement of delivery", and a customizing screen of tagging rules is displayed according to the user's selection. Furthermore, the user can customize, such as adding and deleting keywords associated with the tag, and changing the date, as exemplified in FIG. 8(b). Furthermore, the user can set (fixed condition) to fix the “condition” associated with the tag. A "condition" set as a fixed condition is always retained as a "condition" without being removed when a rule is updated by a user's tagging operation.
Further, as exemplified in FIGS. 9A and 9B, the tagging rule customization screen can be called up when a tag is proposed to the user and also from the tag list that is the search result of the tag. In this way, since the user can check the tagging rules and modify them as necessary, it is possible to generate tagging rules that cannot be handled by updating the tagging rules by the file management device 5 .

FIG. 10 is a flow chart for explaining tagging and tag proposal processing (S20) by the file management device 5. As shown in FIG.
As illustrated in FIG. 10, in step 200 (S200), the acquisition unit 500 acquires image data of the document scanned by the scanner 3. FIG. A common feature extraction unit 502 acquires a data file obtained by performing OCR processing on the image data acquired by the acquisition unit 500 . A common feature extraction unit 502 extracts at least one of character strings, dates, image sizes, and the number of colors used in images as features.
At step 205 ( S<b>205 ), the collation unit 504 collates the feature of the data file with the tagging rules stored in the tagging rule DB 600 .
At step 210 (S210), the matching unit 504 proceeds to S215 if the feature of the data file is matched with all tagging rules, and proceeds to S205 if not matched with all tagging rules. Transition.
At step 215 (S215), the matching unit 504 selects the tagging rule with the highest matching rate with the feature of the data file as a result of the matching.
At step 220 (S220), if the matching rate of the tagging rule selected at S215 is 100%, the process proceeds to S225, and if the matching rate is not 100%, the process proceeds to S235.

At step 225 (S225), the tagging unit 508 gives the data file a tag of a tagging rule with a matching rate of 100%.
At step 230 (S230), the rule updating unit 510 updates and registers the tagging rule. Specifically, among the tagging rules selected in S215, conditions that do not correspond to the features extracted by the common feature extraction unit 502 are deleted. Furthermore, a "condition" is selected based on the score of each condition candidate, a fixed condition, and other condition candidates, the "condition" of the tagging rule is replaced with the selected new "condition", and the tagging rule DB 600 to register.
In step 235 (S235), if the tagging rule selected in S215 and the data file have a matching rate of 50% or more and less than 99% (similar), the tagging unit 508 proceeds to S240 and If the rate is less than or equal to 49%, no tagging is done and the process ends.
At step 240 (S240), the tagging unit 508 proposes that the tag of the tagging rule determined to be similar be given to the data file, and asks the user to decide whether or not to give the tag.

FIG. 11 is a flowchart for explaining the tagging rule updating process (S30) according to the user's response to the tagging proposal. FIG. 11 illustrates a case where the file management device 5 proposes to the user to add the tag "AAA".
As illustrated in FIG. 11, at step 300 (S300), the tagging unit 508 proposes to the user to give the tag "AAA" as illustrated in FIG. Specifically, the tagging unit 508 presents, as the user's response operation, the addition of the tag "AAA", no tagging, and the addition of another tag.
In step 305 (S305), if the user determines that the tag "AAA" is appropriate for the tag proposal by the tagging unit 508, the process proceeds to S310 and does not determine that the tag "AAA" is appropriate. If so, the process proceeds to S320.
At step 310 (S310), the tag assigning unit 508 assigns the tag "AAA" to the data file.
At step 315 (S315), the rule updating unit 510 updates and registers the tagging rule for the tag "AAA". Specifically, the rule update unit 510 selects a “condition” such that the matching rate between the characteristics of the data file and the tagging rule for the tag “AAA” is 100%, and updates the existing “AAA” tag. Replace the "condition" of the attachment rule. The rule update unit 510 does not replace the "condition", but partially deletes the "condition" (for example, relaxes the condition such that the character string matches the three characters to match the two characters) to increase the matching rate. You may do so.

In step 320 (S320), if the user selects to add a tag other than "AAA", here, if the user selects to add the tag "BBB", the process proceeds to S325, and Otherwise, the process proceeds to S340.
At step 325 (S325), the tag assigning unit 508 attaches the tag "BBB" to the data file.
In step 330 (S330), when the user selects "apply tag 'BBB'", the rule updating unit 510 determines that the match rate between the data file characteristics and the tagging rule for the tag 'AAA' is 49. % or less, update the tagging rule for the tag "AAA". Specifically, the rule updating unit 510 selects the “condition” so that the matching rate between the feature of the data file and the tagging rule of the tag “AAA” is 49% or less. Further, the rule updating unit 510 replaces the selected “condition” with the “condition” of the existing “AAA” tagging rule. As a result, the feature of the data file and the tagging rule of the tag "AAA" are not determined to be similar. Also, the rule update unit 510 may reduce the match rate by adding a "condition" (strengthening the condition) instead of replacing the "condition".

At step 335 (S335), the rule updating section 510 updates the tagging rule for the tag "BBB" so that the matching rate between the feature of the data file and the tagging rule for the tag "BBB" is 100%. Specifically, the rule updating unit 510 selects the “condition” so that the matching rate between the feature of the data file and the tagging rule for the tag “BBB” is 100%. Furthermore, the rule update unit 510 replaces the selected “condition” with the “condition” of the existing “BBB” tagging rule. As a result, it is determined that the characteristics of the data file and the tagging rule for the tag "BBB" match. Further, the rule updating unit 510 may increase the match rate by partially deleting the “condition” (relaxing the condition) instead of replacing the “condition”.
In step 340 (S340), when the user selects "do not add tag 'AAA'", the rule updating unit 510 determines that the match rate between the data file characteristics and the tagging rule for the tag 'AAA' is 49. % or less to update the tagging rule for the tag "AAA". More specifically, the rule updating unit 510 selects the “condition” so that the matching rate between the characteristics of the data file and the tagging rule for the tag “AAA” is 49% or less. Then, the rule updating unit 510 replaces the selected “condition” with the “condition” of the existing “AAA” tagging rule. As a result, the feature of the data file and the tagging rule of the tag "AAA" are not determined to be similar. Also, the rule update unit 510 may reduce the match rate by adding a "condition" (strengthening the condition) instead of replacing the "condition".
At step 345 (S345), the rule update unit 510 registers the replaced "condition" in the tagging rule DB 600 as a condition of the tagging rule.

Next, an example of updating the tagging rule in the case where the existing tag "invoice (2018)" given to the document A and the document B is given to the newly tagged document C will be described.
FIG. 13(a) is a diagram representing the tagging rules of "invoice (2018)" and the characteristics of document C, representing keywords, document dates and sizes of documents A, B and C. FIG. As illustrated in FIG. 13A, document C, like documents A and B, does not include the character strings "below" and "Tokyo".
FIG. 13(b) is a table showing scores of keywords of condition candidates.
FIG. 13(c) is an example of updating the tagging rule data.
The rule updating unit 510 selects a keyword to be adopted as the "condition" based on the score of the "condition candidate" included in the document C. FIG. Specifically, as exemplified in FIG. ” to “9”, and the appearance positions of “invoice” and “tax amount” are added. As a result, the total score of "below" and "Tokyo" decreases, and the total score of "invoice" and "tax amount" increases. Therefore, as exemplified in FIG. 13C, the rule updating unit 510 sets the tagging rules generated based on the document A and the document B as keywords "invoice", "amount", "transfer", The tagging rule is updated to give a tag "invoice (2018)" to a data file having "tax amount" and "delivery date", having "2018" as the document date, and having a size of "A4".

Next, an example of tagging rule update when a tagging rule with a matching rate of 100% cannot be generated with only one condition will be described. Specifically, a case will be described in which the user attaches the existing tag "bill" to the newly tagged document D. FIG.
FIG. 14(a) is a diagram showing the tagging rules for "bill" and the features of document D, showing the keywords of document A, document B, document C, and document D, and the document date. As exemplified in FIG. 14(a), document D does not include the character string "invoice" like documents A to C, and documents A to C and newly tagged document D: No common data file characteristics. That is, No. in FIG. 14(c). Document D cannot be given the tag "bill" with only one existing tagging rule of "bill".
FIG. 14(b) is a table showing scores of keywords of condition candidates.
FIG. 14(c) is an example of updating the tagging rule data.
As illustrated in FIGS. 14(b) and 14(c), the score calculation unit 506 recalculates the score of the “condition candidate” keyword, and the rule update unit 510 updates the tagging rule before update ( In addition to No. 1), in order for document D to fall under the tag "invoice", "payment destination" and "payment deadline" with a high total score are selected as "conditions" and tagged with the tag "invoice" Add as a rule. Therefore, the rule updating unit 510 registers the conditions "No. 1" and "No. 2" as the tagging rules for the tag "bill". As a result, if the condition of the tagging rule is "No. 1" or "No. 2", the tag "invoice" is added.

Next, an example of updating the tagging rules when the user rejects the suggestion after tagging is suggested will be described. Specifically, a case will be described in which the user rejects the proposal of the existing tag "AA company_bill" for the newly tagged document G, which is the bill of company BB.
FIG. 15(a) is a diagram showing the tagging rule of "Company AA_invoice" and the characteristics of the newly tagged document G, which is the invoice of Company BB. FIG. 10 is a diagram illustrating tagging rules for "_bill", (c) is a table showing scores of keywords of condition candidates, and (d) is an example of updating tagging rule data.
According to the tagging rule of "Company XX_invoice" illustrated in FIG. 15B, the tag "Company AA_invoice" is proposed to the invoice of Company BB as well.
Therefore, as exemplified in FIG. 15A, the score calculation unit 506 compares the features of the document G for which tagging is rejected with the "candidate conditions" of the tagging rule with the tag "Company AA_bill". do. If there is a “candidate condition” that is not included in the features of document G among the “candidate conditions” of the tagging rule of the tag “company AA_invoice”, the score calculation unit 506 Add gender score. Specifically, as exemplified in FIG. 15C, the score calculation unit 506 calculates the peculiarity of “AA Company”, which is a keyword condition candidate that is included in documents E and F but not included in document G. give a score. As a result of the recalculation of the score by the score calculating unit 506, the rule updating unit 510 sets the tagging rule to the tag name "AA company_bill" and the keyword ""AA company", as illustrated in FIG. 15(d). , “Invoice”, “Payee”, “Payment Due Date”, and “Billed Amount”. As a result, the tagging unit 508 no longer proposes the tag "AA company_bill" to document G, which is the bill of company BB.

As described above, according to the file management device 5, tags can be automatically added to data files acquired from the scanner 3 based on the features of the data files and the tagging rules without user intervention. In addition, the user can review the tagging rules managed by the file management device 5 and correct the tagging rules as necessary. The tagging rules are then used to establish more accurate tagging rules because the tagging rules are updated based on the matching rate between the tagged document and the tagging rules.

In the above embodiment, the file management device 5 attaches tags to the image data read by the scanner 3, but the present invention is not limited to this, and the scanner 3 has the function of the file management device 5. , the image data may be read and tags may be added to the data files. Furthermore, the user terminal 7 may have the function of the file management device 5, and the user terminal 7 may add tags to data files.

1 file management system 3 scanner 5 file management device 50 file management program 500 acquisition unit 502 common feature extraction unit 504 collation unit 506 score calculation unit 508 tag assignment unit 510 rule update unit 600 tag tagging rule database

Claims (9)

a common feature extraction unit for extracting features common to a plurality of data files to which the same tag is attached;
a rule storage unit that associates at least one feature extracted by the common feature extraction unit and one tag that has been assigned to these data files with each other and stores them as an assignment rule;
A file management device comprising: a tag assigning unit that assigns a tag to a newly input data file based on the assigning rule stored in the rule storage unit. The tagging unit searches the newly input data file for the features registered as the granting rules in the rule storage unit, and if any feature is found, the tag associated with this feature to a newly input data file. When part of a character string registered as a feature of a tagging rule is found in a newly input data file, the tagging unit proposes a tag associated with this feature to the user, 3. The file management device according to claim 2, wherein a tag is added according to the operation of . 4. The method according to claim 3, further comprising: a rule update unit that updates the grant rule so that, when the proposed tag is adopted by the user, the features of the newly input data file match the features of the grant rule. File management device. 4. The method according to claim 3, further comprising: a rule updating unit that updates the granting rule so that the features of the newly input data file do not match the features of the granting rule if the proposed tag is not adopted by the user. file management device. 2. The file management device according to claim 1, wherein the common feature extractor extracts at least one of a character string, a date, an image size, and the number of colors used in an image as the feature. The addition rule stored in the rule storage unit includes a plurality of features as determination elements for determining whether or not to add a tag ,
The rule update unit selects features to be registered as determination elements of the given rule based on at least one of appearance frequency, recency, and appearance position, and uniqueness from features common to a plurality of data files. The file management device according to claim 4 or 5. a common feature extraction step in which a computer extracts features common to a plurality of data files to which the same tag is attached;
a rule storage step in which the computer associates at least one feature extracted by the common feature extraction step with one tag assigned to these data files and stores them as an assignment rule;
A file management method comprising: a tagging step in which a computer tags newly input data files based on the granting rules stored in the rule storing step. a common feature extraction step of extracting features common to a plurality of data files to which the same tag is attached;
a rule storage step of correlating at least one feature extracted by the common feature extraction step and one tag assigned to these data files with each other and storing them as an assignment rule;
A program for causing a computer to execute: a tagging step of tagging a newly input data file based on the tagging rule stored in the rule storing step.

Images